Canon EOS 5D MKIV focus question

[fzihlmann]

I am very happy with my MKIV and have an auto focus question. I mostly shoot in “manual select one point AF” by manually selecting one of the 61 AF points on the cameras AF sensor. I also use the Live View “Flexi Zone Single AF” when shooting focus stacking images. Shooting in “Flexi Zone Single AF” the focus area within the frame is significantly larger than in “manual select one point AF”. Is there a difference in focus accuracy between the two focus methods? I know that the “Flexi Zone Single AF” method uses the Dual Pixel technology but I do not know if one pays a penalty in this mode.Anybody out there who has some clarification on this question?

[kvbarkley]

Almost by definition, Live View is more accurate since it is focusing at the sensor. The standard AF uses a separate optical path and focuses at the bottom of the mirror box.

[TCampbell]

The camera uses a completely different focus system when doing normal photography through the viewfinder vs. using movies or using the live-view mode.

For video AF, here's a good tutorial:

When you use auto-focus during either live-view or movie modes, the camera uses the "dual-pixel CMOS AF" ... but keep in mind that this system means that the pixels on the sensor are split into halves and the camera can use phase-detect-like system (a beam splitter seperates light into two halves and they converge if the subject is focused but will be out-of-phase if not focused (and the camera can determine how far out-of-phase and in which direction so that it very rapidly corrects focus).

The difference between the flexi-zone options is whether you want to restrict the active AF points to a specific area.  e.g. should it use all the AF points and look for the nearest focusable object?  Should it restrict focus to a certain area of the frame (so a closer subject outside that selected box will be ignored)?   Should it look for something that resembles a face?  Regardless of behavior... it's doing it all with lots of those Dual-Pixel CMOS AF points.  Canon says it's about 80% of the sensor that has those points.  It's not the same pattern of just 61 points that you see when you use the viewfinder for normal photography.

Back to the normal viewfinder mode....

The AF points drawn on the focus screen when you look through the viewfinder aren't physical representations of the focus sensor array inside the camera.   They are mostly there to show you "where" the focus point is located ... but not it's actual size.

The 5D Mk IV has a number of ways it can use these points.  I'll highlight a few of them.

• Singe Point AF - in this mode the camera uses just a single AF point but it uses the entire point.  This is a nice mode when the subject offers nice contrasty pattern or texture so the camera shouldn't have any trouble focusing and also when it shouldn't be difficult to pick off the correct spot on your subject.
• Spot AF - in this mode, the camera REDUCES the size of the single point.   This mode should ONLY be used if the focused position on your subject has nice contrasty detail so that the camera can easily achieve focus (the system needs contrasty details to focus accurately).   Why reduce the size?  Suppose I am photographing a model and I want to lock focus on the eyeball.  The eyeball and eyelash probably provides lots of contrast.  But I do NOT want the camera to focus on the eye-brow (which is slightly closer) nor the nose, etc.  So by reducing the size of the AF point, I can be sure that the camera will pick off just that one spot and not be distracted by other contrasty things that are not actually at the same distance from the lens.
• Expaded AF - in this mode you'll notice the camera highlights a pattern of 5 focus points... a center point along with a point which is above & below as well as left & right.  The real AF point is that center point.  But the camera is "borrowing" the points above/below/left/right to increase the area of AF.  If a subject has moderately poor contrast then borrowing those extra points increases the effective size of the AF point to make it easier to achieve focus.  
• Surround AF - this is the same as expanded AF (above) except it even throws in the 4 corners.  So now you have a cluster of 9 AF points.  Just remember it's the CENTER point that the camera wants and the rest are being used to enlarge the size.  Again, this is mode helps the camera in situations where focus is too difficult with just single point AF (maybe the subject is moving, etc.) and increasing the effective size of the AF helps the camera achieve focus faster.

There are more modes but I wanted to highlight these four.

You can think of them as "small", "normal", "large", and "extra large" where the "small" allows for the finest control of the precise location of focus on a subject (such as the eyes) with the caveat that it will only likely succeed if that spot offers lots of contrasty details (an eye would qualify as "contrast") AND the subject shouldn't be moving.

And then you can move up through the sizes based on degree of difficulty achieving focus with that extra-large size (surround AF) being used for situations where you're trying to focus on a fast moving subject or a subject with poor contrast (restricting the camera to a tiny tight spot might not offer enough contrast to help the camera achieve focus fast enough).

Here's another Canon video on the topic:

One MAJOR requirement of all of the above modes is that these focus sensors are in the floor of the camera (they are not part of the camera's imaging sensor).  The reflex mirror inside the camera has a semi-transparent region in the middle.  So while it's bouncing the image up and onto the focus screen (what you see when you look through the camera), some light actually passes through that first mirror and hits a tiny secondary mirror hiding behind the first mirror.  That secondary mirror is on a 45º angle so it bounces light down and into the floor of the camera.  

That secondary mirror is responsible for getting the light to the normal auto-focus sensors when using viewfinder mode (normal photography).

When you switch to either live-view mode or video, the reflex mirror has to swing up to clear the light path so that light can reach the sensor.  The secondary mirror folds into the back of the primary mirror and both swing up to the top of the camera (blocking the focus screen -- so if you try to look through the vewfinder it'll just be black).  

At this point, the camera cannot use the normal AF points because no light is being reflected down into them.

When you are using live-view or shooting video, the camera uses the "Dual-Pixel CMOS AF" built into the sensor in order to focus.  

The camera is able to use EITHER the AF sensors on the floor of the camera (when using normal viewfinder photography) -OR- it can use Dual-Pixel CMOS AF (built into the imaging sensor) when you use either live-view or movie modes.   It can never use both at the same time.

[fzihlmann]

Thank you kvbarkley and Tim.

Regarding my question about focus accuracy between the "one point AF" in the 61 point AF system and Flexi Zone Single AF in the Live View dual pixel environment:

Based on your feedback it appears they are equal and perhaps the dual pixel based focus may be even more accurate. I am an extensive user of both focusing sytsems and I selectively use them based on what I am shooting.

[TCampbell]

---

@fzihlmann wrote:

Thank you kvbarkley and Tim.

Regarding my question about focus accuracy between the "one point AF" in the 61 point AF system and Flexi Zone Single AF in the Live View dual pixel environment:

Based on your feedback it appears they are equal and perhaps the dual pixel based focus may be even more accurate. I am an extensive user of both focusing sytsems and I selectively use them based on what I am shooting.

---

I haven't done any performance benchmarks but my "guess" is that the phase-detect AF system (the sensors in the floor of the camera) are probably faster (I don't know this... but I suspect it) because the sensors are physically larger and it's likely easier for them to measure focus accuracy.

But the sensor inside the camera is mounted on adjustable shims (factor adjustable... not user adjustable).  It is critically important that the sensor be level top-to-bottom and left-to-right or you can end up with images that are sharp on one side and soft on the other.  When the factory does this, they also have to calibrate the phase-detect AF sensors.  

"In theory" the distance that light travels as it enters the lens, hits the reflex mirror sending much of the light onto the focus screen and some of the light down to the phase-detect focus sensors is supposed to be identical AND when the reflex mirror swings clear to take a photo, the distance from the point where the light would have bounced down to the focus sensors vs. the actual path of continuing forward to the imaging sensor is also supposed to be identical.  But all of this has to be calibrated.

This means it is technically possible for the sensors on the floor of the camera to believe you have accurate focus when true inspection of the image shows that the image is slightly out of focus.  

For this reason, you camera is equipped with a feature called "Auto Focus Micro-Adjustment" (often abbreviated AFMA or sometimes AF Micro-Adjust).  If you find your camera isn't accurate, you can tweak it yourself (you can even independently do this for each lens you own.)

When you use the imaging sensor itself to focus (as is the case of Dual-Pixel CMOS AF) there's no need to calibrate because the phase-detect sensors on the floor aren't being used.  The focusing system is literally built into the sensor.  In theory you should get bang-on accurate auto-focus every time.

You might be thinking...  why not use the live-view all the time and never use the other system?  

There are a few reasons.

1)  Live-view draws more power.   Batteries wont last as long if you do this.

2)  The sensor is continuously active and that causes it to heat up.  A warmer sensor generates more "noise" in the image.  You'll have a cleaner image if you try to keep the sensor from getting too warm.

3)  It's more stable to do hand-held photography through the viewfinder because the camera is in close to your body and supported from underneath (arms tucked in close to your stomach) and less stable to hold the camera at a distance where your eyes can see the view-screen.

I'm not certain if it's still true that the phase-detect system is faster to auto-focus.  That used to be true, but the dual-pixel CMOS AF keeps getting better with each new generation and I haven't seen any tests on the current generation.

Use both systems as/when they fit your needs.  I don't think I'd use either exclusively (there's a reason your camera has both systems.)

[fzihlmann]

I use Live View Dual Pixel focusing for certain landscape situation, especially if I do focus stacking. For wildlife, I exclusively use the 61 point AF system.

I do check the auto focus accuracy of my lenses with the LensAlign tool but to this point I did not have to make an "Auto Focus Micro-Adjustment".